To understand how genome sequences relate to the biology of the organism, correct annotation of gene function is essential. There are many computational approaches to assigning a putative gene function based on similarity to known activities. However, as the evolutionary distances extend and the number

Gallery of structures solved by JCSG

of putative homologs expands, the reliability of such unvalidated predictions becomes suspect. Likewise, while enzyme classes and broad activities may be predicted with some certainty, much less can be inferred regarding substrate specificity. This specificity defines the role of that enzyme at a functional level and understanding enzyme selectivity is essential to understanding the enzyme in the biology of the organism.

In collaboration with Andrei Osterman (The Burnham Institute) and Bernhard Geierstanger (GNF) we are undertaking an approach to such characterization which can be applied generally and in a high-throughput fashion to proteins with unknown or putative functions. 

Our approach comprises:  a bioinformatic platform to analyze gene relationships and propose putative functions for testing, utilization of an existing high-throughput platform for producing thousands of proteins in milligram quantities for analysis, ligand screening technologies combined with metabolite-focused compound libraries to define specificity profiles of enzymes, and focused experimental validation of proposed functions of key target genes and pathways.